Construction,validation, and application of nocturnal pollen transport networks in an agro‐ecosystem: a comparison using light microscopy and DNA metabarcoding

1. Moths are globally relevant as pollinators but nocturnal pollination remains poorly understood. Plant–pollinator interaction networks are traditionally constructed using either flower‐visitor observations or pollen‐transport detection using microscopy. Recent studies have shown the potential of DNA metabarcoding for detecting and identifying pollen‐transport interactions. However, no study has directly compared the realised observations of pollen‐transport networks between DNA metabarcoding and conventional light microscopy. 2. Using matched samples of nocturnal moths, we constructed pollen‐transport networks using two methods: light microscopy and DNA metabarcoding. Focussing on the feeding mouthparts of moths, we developed and provide reproducible methods for merging DNA metabarcoding and ecological network analysis to better understand species interactions. 3. DNA metabarcoding detected pollen on more individual moths, and detected multiple pollen types on more individuals than microscopy, although the average number of pollen types per individual was unchanged. However, after aggregating individuals of each species, metabarcoding detected more interactions per moth species. Pollen‐transport network metrics differed between methods because of variation in the ability of each to detect multiple pollen types per moth and to separate morphologically similar or related pollen. We detected unexpected but plausible moth–plant interactions with metabarcoding, revealing new detail about nocturnal pollination systems. 4. The nocturnal pollination networks observed using metabarcoding and microscopy were similar yet distinct, with implications for network ecologists. Comparisons between networks constructed using metabarcoding and traditional methods should therefore be treated with caution. Nevertheless, the potential applications of metabarcoding for studying plant–pollinator interaction networks are encouraging, especially when investigating understudied pollinators such as moths.     

Flower‐visitor networks only partially predict the function of pollen transport by bees

Mutualistic networks display distinct structural and organizational features such as nestedness, power‐law degree distribution and asymmetric dependencies. Attention is now focused on how these structural properties influence network function. Most plant‐pollinator networks are constructed using records of animals contacting flowers, which is based on the assumption that all visitors to flowers are pollinators; however, animals may visit flowers as nectar robbers, florivores, or to prey upon other visitors. To differentiate potential pollinator interactions from other interaction types, we examined individual bees that had visited flowers to detect if they carried pollen. Using these data, we constructed visitation and pollen‐transport networks for a spinifex‐dominated arid zone grassland. To determine how the structure of the visitation network reflects pollen transport, we compared the two networks using a null model approach to account for differences in network size. Differences in number of species, nestedness and connectance observed between the visitation and pollen‐transport networks were within expected ranges generated under the null model. The pollen‐transport network was more specialized, had lower interaction evenness, and fewer links compared to the visitation network. Almost half the number of species of the visitation network participated in the pollen‐transport network, and one‐third of unique visitation interactions resulted in pollen transport, highlighting that visitation does not always result in pollination. Floral visitor data indicate potential pollen transporters, but inferring pollination function from visitation networks needs to be performed cautiously as pollen transport resulted from both common and rare interactions, and depended on visitor identity. Although visitation and pollen‐transport networks are structurally similar, the function of all species cannot be predicted from the visitation network alone. Considering pollen transport in visitation networks is a simple first step towards determining pollinators from non‐pollinators. This is fundamental for understanding how network structure relates to network function.     

Observing frugivores or collecting scats: a method comparison to construct quantitative seed dispersal networks

Mutualistic interactions form the basis for many ecological processes and are often analyzed within the framework of ecological networks. These interactions can be sampled with a range of methods and first analyses of pollination networks sampled with different methods showed differences in common network metrics. However, it is yet unknown if metrics of seed dispersal networks are similarly affected by the sampling method and if different methods detect a complementary set of frugivores. This is necessary to better understand the (dis-)advantages of each method and to identify the role of each frugivore for the seed dispersal process. Here, we compare seed removal networks based on the observation of 2189 frugivore visits on ten focal plant species with seed deposition networks constructed by DNA barcoding of plant seeds in 3094 frugivore scats. We were interested in whether both methods identify the same disperser species and if species-level network metrics of plant species were correlated between network types. Both methods identified the same avian super-generalist frugivores, which accounted for the highest number of dispersed seeds. However, only with DNA barcoding, we detected elusive but frequent mammalian seed dispersers. The overall networks created by both methods were congruent but the plant species' degree, their interaction frequency and their specialization index (d′) differed. Our study suggests that DNA barcoding of defecated and regurgitated seeds can be used to construct quantitative seed deposition networks similar to those constructed by focal observations. To improve the overall completeness of seed dispersal networks it might be useful to combine both methods to detect interactions by both birds and mammals. Most importantly, the DNA barcoding method provides information on the post-dispersal stage and thus on the qualitative contribution of each frugivore for the plant community thereby linking species interactions to regeneration dynamics of fleshy-fruited plant species.     

Invariant antagonistic network structure despite high spatial and temporal turnover of interactions

Recent work has suggested that emergent ecological network structure exhibits very little spatial or temporal variance despite changes in community composition. However, the changes in network interactions associated with turnover in community composition have seldom been assessed. Here we examine whether changes in ecological networks are best detected by standard emergent network metrics or by assessing internal network changes (i.e. interaction and composition turnover). To eliminate possible spatial or phylogenetic effects, that in large‐scale studies may obscure mechanisms structuring networks and interactions, we sampled multiple antagonistic (plant–herbivore) networks for a single diverse plant family (the Restionaceae) in the hyperdiverse Cape Floristic Region. These are the first plant–herbivore networks constructed for this global biodiversity hotspot. We found invariant emergent network structure despite considerable changes in insect and plant composition across communities over time and space. In contrast, there was high interaction turnover between networks. Seasonally, this was driven by turnover in insect species and insect host switching. Spatially, this was driven by simultaneous turnover in plant and insect species, suggesting that many insects are host specific or that both groups exhibit parallel responses to environmental gradients. Spatial interaction turnover was also driven by turnover in plants, showing that many insects can utilise multiple (possibly closely related) hosts and this may create divergent selection gradients that promote insect speciation. Thus we show highly variable interaction fidelity, despite invariant emergent network structure. We suggest that evaluating internal network changes may be more effective at elucidating the processes structuring networks, and many fine‐scale changes may be obscured when only calculating emergent network metrics.     

Seeing is believing? Comparing plant–herbivore networks constructed by field co‐occurrence and DNA barcoding methods for gaining insights into network structures

Plant–herbivore interaction networks provide information about community organization. Two methods are currently used to document pairwise interactions among plants and insect herbivores. One is the traditional method that collects plant–herbivore interaction data by field observation of insect occurrence on host plants. The other is the increasing application of newly developed molecular techniques based on DNA barcodes to the analysis of gut contents. The second method is more appealing because it documents realized interactions. To construct complete interaction networks, each technique of network construction is urgent to be assessed. We addressed this question by comparing the effectiveness and reliability of the two methods in constructing plant–Lepidoptera larval network in a 50 ha subtropical forest in China. Our results showed that the accuracy of diet identification by observation method increased with the number of observed insect occurrences on food plants. In contrast, the molecular method using three plant DNA markers were able to identify food residues for 35.6% larvae and correctly resolved 77.3% plant (diet) species. Network analysis showed molecular networks had threefold more unique host plant species but fewer links than the traditional networks had. The molecular method detected plants that were not sampled by the traditional method, for example, bamboos, bryophytes and lianas in the diets of insect herbivores. The two networks also possessed significantly different structural properties. Our study indicates the traditional observation of co‐occurrence is inadequate, while molecular method can provide higher species resolution of ecological interactions.     

The importance of interannual variation and bottom–up nitrogen enrichment for plant–pollinator networks

Striking changes in food web structure occur with alterations in resource supply. Like predator–prey interactions, many mutualisms are also consumer–resource interactions. However, no studies have explored how the structure of plant–pollinator networks may be affected by nutrient enrichment. For three years, we enriched plots of subalpine plant communities with nitrogen and observed subsequent effects on plant–pollinator network structure. Although nitrogen enrichment affects floral abundance and rates of pollinator visitation, we found no effects of nitrogen enrichment on the core group of generalist plants and pollinators or on plant–pollinator network structure parameters, such as network topology (the identity and frequency of interactions) and the degree of nestedness. However, individual plant and pollinator taxa were packed into the nested networks differently among nitrogen treatments. In particular, pollinators visited different numbers and types of plants in the nested networks, suggesting weak, widespread effects of nitrogen addition on individual taxa. Independent of nitrogen enrichment, there were large interannual differences in network structure and interactions, due to species turnover among years and flexibility in interacting with new partners. These data suggest that the community structure of small‐scale mutualistic networks may be relatively robust to short‐term bottom–up changes in the resource supply, but sensitive to variation in the opportunistic behavior and turnover of plant and pollinator species among years.     

Plant–mycorrhizal fungus co‐occurrence network lacks substantial structure

The interactions between plants and arbuscular mycorrhizal fungi (AMF) maintain a crucial link between macroscopic organisms and the soil microbial world. These interactions are of extreme importance for the diversity of plant communities and ecosystem functioning. Despite this importance, only recently has the structure of plant–AMF interaction networks been studied. These recent studies, which used genetic data, suggest that these networks are highly structured, very similar to plant–animal mutualistic networks. However, the assembly process of plant–AMF communities is still largely unknown, and an important feature of plant–AMF interactions has not been incorporated: they occur at an extremely localized scale. Studying plant–AMF networks in a spatial context seems therefore a crucial step. This paper studies a plant–AMF spatial co‐occurrence network using novel methodology based on information theory and a unique set of spatially explicit species‐level data. We apply three null models of which only one accounts for spatial effects. We find that the data show substantial departures from null expectations for the two non‐spatial null models. However, for the null model considering spatial effects, there are few significant co‐occurrences compared with the other two null models. Thus, plant–AMF spatial co‐occurrences seem to be mostly explained by stochasticity, with a small role for other factors related to plant–AMF specialization. Furthermore, we find that the network is not significantly nested or modular. We conclude that this plant–AMF spatial co‐occurrence network lacks substantial structure and, therefore, plants and AMF species do not track each other over space. Thus, random encounters seem more important in the first step of the assembly of plant–AMF communities. Synthesis The symbiotic interaction between plants and arbuscular mycorrhizal fungi (AMF) is crucial for ecosystem functioning. However, the factors affecting the assembly of plant‐AMF communities are poorly understood. An important factor of the assembly of plant‐AMF communities has been overlooked: plant‐AMF interactions occur at a localized spatial scale. Our study investigated the importance of space in the structure of plant‐AMF communities. We studied a plant‐AMF spatial co‐occurrence network using a unique set of spatially explicit data and applied three null models. We found that plant‐AMF spatial co‐occurrences seem to be mostly explained by stochasticity. In particular, our study shows that this plant‐AMF spatial co‐occurrence network lacks substantial structure and, therefore, plants and AMF species do not track each other over space. Thus, random encounters seem to drive the assembly of plant‐AMF communities.     

The role of the endemic and critically endangered Colorful Puffleg Eriocnemis mirabilis in plant‐hummingbird networks of the Colombian Andes

Ecological network approaches may contribute to conservation practices by quantifying within‐community importance of species. In mutualistic plant‐pollinator systems, such networks reflect potential pollination of the plants and a considerable portion of the energy consumption by the pollinators, two key components for each party. Here, we used two different sampling approaches to describe mutualistic plant‐hummingbird networks from a cloud forest in the Colombian Western Andes, home to the Colorful Puffleg Eriocnemis mirabilis, an endemic and critically endangered hummingbird. We contrast networks between two localities (a protected area inside a National park vs. its buffer zone) and across sampling methods (floral visitation vs. pollen loads) to assess how the network structure and the importance of each hummingbird species within the networks may change. Visitation networks were characterized as having higher sampling completeness, yet pollen load network recorded more pollen types than plant species recorded by visitation. Irrespective of the sampling methods, the Colorful Puffleg was one of the most important hummingbird species in the network within the protected area inside the National park, but not in the buffer zone. Moreover, most species‐level network indices were related to hummingbirds’ abundance. This suggests that conservation initiatives aimed at the endangered Colorful Puffleg may both help on the survival of this endangered hummingbird, as well as on maintaining its key role in the mutualistic interaction network inside the National Park. Our study illustrates how conservation practitioners could assess the local importance of endangered species using interaction network approaches.     

Do asynchronies in extinction debt affect the structure of trophic networks? A case study of antagonistic butterfly larvae–plant networks

Habitat loss and fragmentation affect species richness in fragmented habitats and can lead to immediate or time‐delayed species extinctions. Asynchronies in extinction and extinction debt between interacting species may have severe effects on ecological networks. However, these effects remain largely unknown. We evaluated the effects of habitat patch and landscape changes on antagonistic butterfly larvae–plant trophic networks in Mediterranean grasslands in which previous studies had shown the existence of extinction debt in plants but not in butterflies. We sampled current species richness of habitat‐specialist and generalist butterflies and vascular plants in 26 grasslands. We assessed the direct effects of historical and current patch and landscape characteristics on species richness and on butterfly larvae–plant trophic network metrics and robustness. Although positive species‐ and interactions–area relationships were found in all networks, structure and robustness was only affected by patch and landscape changes in networks involving the subset of butterfly specialists. Larger patches had more species (butterflies and host plants) and interactions but also more compartments, which decreased network connectance but increased network stability. Moreover, most likely due to the rescue effect, patch connectivity increased host‐plant species (but not butterfly) richness and total links, and network robustness in specialist networks. On the other hand, patch area loss decreased robustness in specialist butterfly larvae–plant networks and made them more prone to collapse against host plant extinctions. Finally, in all butterfly larvae–plant networks we also detected a past patch and landscape effect on network asymmetry, which indicates that there were different extinction rates and extinction debts for butterflies and host plants. We conclude that asynchronies in extinction and extinction debt in butterfly–plant networks provoked by patch and landscape changes caused changes in species richness and network links in all networks, as well as changes in network structure and robustness in specialist networks.     

Effect of agricultural land‐use change on the structure of a temperate forest ant–plant interaction network

Studies on the responses of ant–plant interactions to land‐use change have mainly focused on tropical habitats, usually without considering the impacts on the structure of interaction networks. Here we show that land‐use modifies the structure of the ant–plant interaction networks in a temperate habitat. Ant–plant interactions and plant diversity were recorded in an oak forest and agricultural land in central Mexico. We registered five ant species in the oak forest, and four ant species in the agricultural land. Plant diversity was higher in the agricultural land than in the oak forest. In the ant–plant networks of both sites, our results showed a higher dependence of ants on the plants on which they feed than vice versa, and the ants Formica spp. and the plants Barkleyanthus salicifolius were the species with the most strength and greatest influence in the network structure. The ant–plant network in the oak forest showed a nested structure. However, the network at the agricultural land site showed non‐nestedness; the identity of both ants and plants with the highest values of specialization was different and the number of ant species in the network was decreased, but the number of plant species with which they interacted significantly increased. Both ant–plant networks were equally tolerant to simulated extinction of individual species. We conclude that temperate forest ant–plant networks can be inherently fragile and susceptible to the effects of agricultural land‐use change, not on the number of interacting species but on their identity.     

Morphological and Spatio‐Temporal Mismatches Shape a Neotropical Savanna Plant‐Hummingbird Network

Complex networks of species interactions might be determined by species traits but also by simple chance meetings governed by species abundances. Although the idea that species traits structure mutualistic networks is appealing, most studies have found abundance to be a major structuring mechanism underlying interaction frequencies. With a well‐resolved plant–hummingbird interaction network from the Neotropical savanna in Brazil, we asked whether species morphology, phenology, nectar availability and habitat occupancy and/or abundance best predicted the frequency of interactions. For this, we constructed interaction probability matrices and compared them to the observed plant‐hummingbird matrix through a likelihood approach. Furthermore, a recently proposed modularity algorithm for weighted bipartite networks was employed to evaluate whether these factors also scale‐up to the formation of modules in the network. Interaction frequencies were best predicted by species morphology, phenology and habitat occupancy, while species abundances and nectar availability performed poorly. The plant–hummingbird network was modular, and modules were associated to morphological specialization and habitat occupancy. Our findings highlight the importance of traits as determinants of interaction frequencies and network structure, corroborating the results of a previous study on a plant–hummingbird network from the Brazilian Atlantic Forest. Thus, we propose that traits matter more in tropical plant–hummingbird networks than in less specialized systems. To test the generality of this hypothesis, future research could employ geographic or taxonomic cross‐system comparisons contrasting networks with known differences in level of specialization.     

Spatial structure of an individual‐based plant–pollinator network

The influence of space on the structure (e.g. modularity) of complex ecological networks remains largely unknown. Here, we sampled an individual‐based plant–pollinator network by following the movements and flower visits of marked bumblebee individuals within a population of thistle plants for which the identities and spatial locations of stems were mapped in a 50 × 50 m study plot. The plant–pollinator network was dominated by parasitic male bumblebees and had a significantly modular structure, with four identified modules being clearly separated in space. This indicated that individual flower visitors opted for the fine‐scale division of resources, even within a local site. However, spatial mapping of network modules and movements of bumblebee individuals also showed an overlap in the dense center of the plant patch. Model selection based on Akaike information criterion with traits as predictor variables revealed that thistle stems with high numbers of flower heads and many close neighbours were particularly important for connecting individuals within the modules. In contrast, tall plants and those near the patch center were crucial for connecting the different modules to each other. This demonstrated that individual‐based plant–pollinator networks are influenced by both the spatial structure of plant populations and individual‐specific plant traits. Additionally, bumblebee individuals with long observation times were important for both the connectivity between and within modules. The latter suggests that bumblebee individuals will still show locally restricted movements within sub‐patches of plant populations even if they are observed over a prolonged time period. Our individual‐based and animal‐centered approach of sampling ecological networks opens up new avenues for incorporating foraging behaviour and intra‐specific trait variation into analyses of plant–animal interactions across space.     

The effects of data collection and observation methods on uncertainty of social networks in wild primates

In social species, network centralities of group members shape social transmission and other social phenomena. Different factors have been found to influence the measurement of social networks, such as data collection and observation methods. In this study, we collected data on adults and juveniles and examined the effect of data collection method (ad libitum sampling vs. focal animal sampling) and observation method (interaction—grooming; play—vs. association—arm-length; 2 m; 5 m proximities—) on social networks in wild vervet monkeys. First, we showed using a bootstrapping method, that uncertainty of ad libitum grooming and play matrices were lesser than uncertainty of focal matrices. Nevertheless, grooming and play networks constructed from ad libitum and focal animal sampling were very similar and highly correlated. We improved the certainty of both grooming and play networks by pooling focal and ad libitum matrices. Second, we reported a high correlation between the proximity arm-length network and the focal grooming one making an arm-length proximity network a reasonable proxy for a grooming one in vervet monkeys. However, we did not find such a correlation between proximity networks and the play one. Studying the effects of methodological issues as data collection and observation methods can help improve understanding of what shapes social networks and which data collection method to choose to study sociality.     

Structure and dynamics of pollination networks: the role of alien plants

Research on ecological communities, and plant–pollinator mutualistic networks in particular, has increasingly benefited from the theory and tools of complexity science. Nevertheless, up to now there have been few attempts to investigate the interplay between the structure of real pollination networks and their dynamics. This study is one of the first contributions to explore this issue. Biological invasions, of major concern for conservation, are also poorly understood from the perspective of complex ecological networks. In this paper we assess the role that established alien species play within a host community by analyzing the temporal changes in structural network properties driven by the removal of non‐native plants. Three topological measures have been used to represent the most relevant structural properties for the stability of ecological networks: degree distribution, nestedness, and modularity. Therefore, we investigate for a detailed pollination network, 1) how its dynamics, represented as changes in species abundances, affect the evolution of its structure, 2) how topology relates to dynamics focusing on long‐term species persistence; and 3) how both structure and dynamics are affected by the removal of alien plant species. Network dynamics were simulated by means of a stochastic metacommunity model. Our results showed that established alien plants are important for the persistence of the pollination network and for the maintenance of its structure. Removal of alien plants decreased the likelihood of species persistence. On the other hand, both the full network and the subset native network tended to lose their structure through time. Nevertheless, the structure of the full network was better preserved than the structure of the network without alien plants. Temporal topological shifts were evident in terms of degree distribution, nestedness, and modularity. However the effects of removing alien plants were more pronounced for degree distribution and modularity of the network. Therefore, elimination of alien plants affected the evolution of the architecture of the interaction web, which was closely related to the higher species loss found in the network where alien plants were removed.     

Reshaping our understanding of species’ roles in landscape‐scale networks

In network ecology, landscape‐scale processes are often overlooked, yet there is increasing evidence that species and interactions spill over between habitats, calling for further study of interhabitat dependencies. Here, we investigate how species connect a mosaic of habitats based on the spatial variation of their mutualistic and antagonistic interactions using two multilayer networks, combining pollination, herbivory and parasitism in the UK and New Zealand. Developing novel methods of network analysis for landscape‐scale ecological networks, we discovered that few plant and pollinator species acted as connectors or hubs, both within and among habitats, whereas herbivores and parasitoids typically have more peripheral network roles. Insect species’ roles depend on factors other than just the abundance of taxa in the lower trophic level, exemplified by larger Hymenoptera connecting networks of different habitats and insects relying on different resources across different habitats. Our findings provide a broader perspective for landscape‐scale management and ecological community conservation.     

Volatility of network indices due to undersampling of intraspecific variation in plant–insect interactions

Appropriate sampling effort of interaction networks is necessary to extract robust indices describing the structure of species interactions. Here we show that time-invariant variation in the composition and diversity of interaction partners of plant individuals of the same species explains volatility in aggregate network statistics due to undersampling. Within a multi-species pollinator–plant interaction matrix, we replaced the interactions observed on multiple individuals of a single plant species (Sinapis arvensis, pooled interactions) with the plant–insect interactions observed on a single plant individual. In the resampling approach, we considered the interactions of 1 to 84 S. arvensis individuals in different combinations. For each resampled network, several commonly applied aggregated statistics were calculated to test how intraspecific variation affects the properties of a multi-species network. Our results showed that aggregate statistics are sensitive towards qualitative and quantitative intraspecific variation of flower–visitor interactions within a multi-species network, which may affect the ecological interpretation about the properties of a community. These findings challenge the robustness of commonly applied network indices, confirm the urge for a sufficient and representative sampling of interactions, and emphasize the significance of intraspecific variation in the context of communities and networks.     

Diversity in a complex ecological network with two interaction types

Most studies on ecological networks consider only a single interaction type (e.g. competitive, predatory or mutualistic), and try to developrules for system stability based exclusively on properties of this interaction type. However, the stability of ecological networks may be more dependent on the way different interaction types are combined in real communities. To address this issue, we start by compiling an ecological network in the Doñana Biological Reserve, southern Spain, with 390 species and 798 mu-tualistic and antagonistic interactions. We characterize network structure by looking at how mutualistic and antagonistic interactions are combined across all plant species. Both the ratio of mutualistic to antagonistic interactions per plant, and the number of basic modules with an antagonistic and a mutualistic interaction are very heterogeneous across plant species, with a few plant species showing very high values for these parameters. To assess the implications of these network patterns on species diversity, we study analytically and by simulation a model of this ecological network. We find that the observed correlation between strong interaction strengths and high mutualistic to antagonistic ratios in a few plant species significantly increases community diversity. Thus, to predict the persistence of biodiversity we need to understand how interaction strength and the architecture of ecological networks with different interaction types are combined.